{"id":35280825,"url":"https://github.com/petroniocandido/nde_tsc","last_synced_at":"2026-04-24T20:03:40.347Z","repository":{"id":195764564,"uuid":"693614226","full_name":"petroniocandido/nde_tsc","owner":"petroniocandido","description":"Neural Density Estimation for Time Series Classification ","archived":false,"fork":false,"pushed_at":"2023-09-20T17:45:27.000Z","size":96,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2023-09-21T03:53:39.125Z","etag":null,"topics":["autoencoder","deep-learning","density-estimation","locality-sensitive-hashing","neural-density-estimation","neural-networks","pytorch","self-organizing-map","siamese-network","time-series-classification"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/petroniocandido.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null}},"created_at":"2023-09-19T11:27:11.000Z","updated_at":"2023-09-19T12:29:25.000Z","dependencies_parsed_at":"2023-09-19T13:44:26.008Z","dependency_job_id":null,"html_url":"https://github.com/petroniocandido/nde_tsc","commit_stats":null,"previous_names":["petroniocandido/nde_tsc"],"tags_count":0,"template":null,"template_full_name":null,"purl":"pkg:github/petroniocandido/nde_tsc","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/petroniocandido%2Fnde_tsc","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/petroniocandido%2Fnde_tsc/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/petroniocandido%2Fnde_tsc/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/petroniocandido%2Fnde_tsc/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/petroniocandido","download_url":"https://codeload.github.com/petroniocandido/nde_tsc/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/petroniocandido%2Fnde_tsc/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32238748,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-24T13:21:15.438Z","status":"ssl_error","status_checked_at":"2026-04-24T13:21:15.005Z","response_time":64,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["autoencoder","deep-learning","density-estimation","locality-sensitive-hashing","neural-density-estimation","neural-networks","pytorch","self-organizing-map","siamese-network","time-series-classification"],"created_at":"2025-12-30T14:38:24.842Z","updated_at":"2026-04-24T20:03:40.342Z","avatar_url":"https://github.com/petroniocandido.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Self Organized Neural Density Estimation for Complex Time Series Classification\n\n\n[![made-with-python](https://img.shields.io/badge/Made%20with-Python-1f425f.svg)](https://www.python.org/)  [![made-with-latex](https://img.shields.io/badge/Made%20with-LaTeX-1f425f.svg)](https://www.latex-project.org/) ![PyTorch](https://img.shields.io/badge/PyTorch-%23EE4C2C.svg?style=for-the-badge\u0026logo=PyTorch\u0026logoColor=white)\n\n\n\n* Petrônio C.  L. Silva  \u003cspan itemscope itemtype=\"https://schema.org/Person\"\u003e\u003ca itemprop=\"sameAs\" content=\"https://orcid.org/0000-0002-1202-2552\" href=\"https://orcid.org/0000-0002-1202-2552\" target=\"orcid.widget\" rel=\"noopener noreferrer\" style=\"vertical-align:top;\"\u003e\u003cimg src=\"https://orcid.org/sites/default/files/images/orcid_16x16.png\" style=\"width:1em;margin-right:.5em;\" alt=\"ORCID iD icon\"\u003e\u003c/a\u003e\u003c/span\u003e\n* Omid Orang  \u003cspan itemscope itemtype=\"https://schema.org/Person\"\u003e\u003ca itemprop=\"sameAs\" content=\"https://orcid.org/0000-0002-4077-3775\" href=\"https://orcid.org/0000-0002-4077-3775\" target=\"orcid.widget\" rel=\"noopener noreferrer\" style=\"vertical-align:top;\"\u003e\u003cimg src=\"https://orcid.org/sites/default/files/images/orcid_16x16.png\" style=\"width:1em;margin-right:.5em;\" alt=\"ORCID iD icon\"\u003e\u003c/a\u003e\u003c/span\u003e\n* Felipe A. R. da Silva  \u003cspan itemscope itemtype=\"https://schema.org/Person\"\u003e\u003ca itemprop=\"sameAs\" content=\"https://orcid.org/0000-0003-4567-8504\" href=\"https://orcid.org/0000-0003-4567-8504\" target=\"orcid.widget\" rel=\"noopener noreferrer\" style=\"vertical-align:top;\"\u003e\u003cimg src=\"https://orcid.org/sites/default/files/images/orcid_16x16.png\" style=\"width:1em;margin-right:.5em;\" alt=\"ORCID iD icon\"\u003e\u003c/a\u003e\u003c/span\u003e\n* Fabricio J. E. Costa\n* Frederico G. Guimarães \u003cspan itemscope itemtype=\"https://schema.org/Person\"\u003e\u003ca itemprop=\"sameAs\" content=\"https://orcid.org/0000-0001-9238-8839\" href=\"https://orcid.org/0000-0001-9238-8839\" target=\"orcid.widget\" rel=\"noopener noreferrer\" style=\"vertical-align:top;\"\u003e\u003cimg src=\"https://orcid.org/sites/default/files/images/orcid_16x16.png\" style=\"width:1em;margin-right:.5em;\" alt=\"ORCID iD icon\"\u003e\u003c/a\u003e\u003c/span\u003e\n\nIn case you have any questions, do not hesitate in contact us using the following e-mail: petronio.candido@ifnmg.edu.br\n\n\nThe Time Series Classification is posed such that given a set of $M$ instances of time series $Y = \\{ Y_1, \\ldots, Y_M \\}$, where all instances are IID and have the same number $T$ of samples $y(t) \\in \\mathbb{R}^N, \\; \\forall t=1\\ldots T$, with $N$ attributes, and a discrete set of labels $C = \\{0, \\ldots, k\\}$, each time series $Y_i$ should be associated with a label $j \\in C$.\n\nThe present proposal aims to investigate methods for computing both the unconditional probability distribution $P(Y_i)$ and the conditional probability distribution $P(Y_i | c)$, $\\forall c \\in C$ and $Y_i \\in Y$. Representing the data space in terms of the $[0,1]$ probability space is a challenging task due to the high dimensionality of $Y$, with the order of $R^N \\times T$ dimensions.\n\nTo allow the representation o $Y$ in probability space, the present research proposes an approach based in Neural Networks in two layers: a) the embedding layer implemented using AutoEncoders and b) the density layer implemented using Self Organizing Maps and k-Nearest-Neighbors.\n\nThe embedding layer aims to learn a function $f_\\mathcal{E}: Y \\rightarrow \\mathcal{E}$ that maps the original data space $Y \\in \\mathbb{R}^{T \\times N}$ to a smaller dimensional space $\\mathcal{E} \\in \\mathbb{R}^D$ dimensions, where $D \\ll T \\times N$. The embedding space $\\mathcal{E}$ should keep, as much as possible, the properties and topoloty of the original space $Y$ such as the distances $d(Y_i, Y_j)$ of each pair of instances $Y_i, Y_j \\in Y$, where $d: Y \\times Y \\rightarrow \\mathbb{R}^+$ is a distance function.\n\nThe density layer aims to learn a map $W$ that represent the topology of the embedding space using a square $L \\times A$ grid of cells $w_{i,j} \\in \\mathbb{R}^D$, where each cell $w_{i,j} \\in W$ represents a prototype or representative value of a  region of the $\\mathcal{E}$ space.\n\nOur proposal learns two more discrete maps, $P$ and $P_k$, where $P[w_{i,j}]$ represents the unconditional probability mass of a the cell $w_{i,j}$ in $\\mathcal{E}$ space and $P_k[w_{i,j}]$ represents the the conditional probability mass of a the cell $w_{i,j}$ in $\\mathcal{E}$ space given the class label $k \\in C$.\n\nWith the maps $W$, $P$ and $P_k$ it is possible to calculate the unconditional $P(Y_i)$ and conditional probabilities $P(Y_i|k)$ for any continuous time series in $Y$, by embedding $Y_i$ such that $e_i = f_\\mathcal{E}(Y_i)$, finding the k-nearest-neighbors of $e_i$ using the square map $W$ and then interpolate its probabilities weighted by their distances.\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpetroniocandido%2Fnde_tsc","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fpetroniocandido%2Fnde_tsc","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpetroniocandido%2Fnde_tsc/lists"}